You're right, some of these lines aren't doing what you think. next immediately breaks out of the current while loop and restarts the loop at the beginning, so the following line is never executed. That means it never tells you that $replace is undefined, so what you're trying to do with it won't work.

In general, when your text is arranged in multi-line chunks, it works best to process it that way rather than line-by-line. You can use the input record separator variable $/ to split your input file on something that ends each chunk. That way you don't have to dance around with saving each line long enough to look at the next line and see if something needs to be done to the previous line before outputting it.

Here's an example of what I'm talking about; you can adjust it to use your input and output files. It reads in each chunk, looks to see if it has the piece you want to copy, and copies it into the other spot if it's there.

#!/usr/bin/env perl
use strict; use warnings;

$/ = '/Hit_accession>';

while(<DATA>){
  if( m{<Hit_def>(.+?) odorant-bin} ){
    my $match = $1;
    s{(<Hit_id>).+?(</Hit_id>)}{$1$match$2};
  }
  print $_;
}

__DATA__
<Hit_num>6</Hit_num>
<Hit_id>gnl|BL_ORD_ID|3665984</Hit_id>
<Hit_def>gi|158703516|gb|ABW77886.1| odorant-bin...
<Hit_accession>3665984</Hit_accession>
<Hit_num>7</Hit_num>
<Hit_id>gnl|BL_ORE_ID|3665984</Hit_id>
<Hit_def>gi|153326716|gb|ABF88997.2| odorant-bin...
<Hit_accession>3665984</Hit_accession>
[download]